You Can Count on Crochet Buddy
“Bodystorming” Stitch Tracking & Counting
The Novel Coronavirus Pandemic has lead to an increase in the number of people picking up crafting in order to fend off the boredom while sheltering in place. I, personally, picked up crochet as a pandemic hobby. The idea for Crochet Buddy stemmed off of my own struggles with crochet, namely: counting stitches, and maintaining consistent tensions while pulling yarn.
After creating the Analog Prototype for Crochet Buddy and the Onshape to create a 3D model of Crochet Buddy, which tested Crochet Buddy’s ergonomic handle design and the placement of the stitch counting camera, and it was time to test what considerations will need to be taken into account when crafting the algorithm for stitch tracking and counting using the footage from the camera on the ergonomic handle.
DESIGN
The Crochet Buddy is a fairly simple IoT device with a large impact on those who crochet. The physical device has three main components: an ergonomic handle, a camera facing the crochet hook’s neck, which counts the number of stitches and tracks which type of stitches are being made using a machine learning algorithm. The hook also has a pressure sensitive hook, which tracks the amount of force exerted each time the user pulls for the stitch. The hook will also vibrate when the user should stop pulling on the stitch, thus enforcing a consistent pull, making the yarn-craft more consistent.
PROTOTYPE
Using my iPad and my analog crochet buddy prototype, I set up a little station to mimic what the camera would see if it were actually attached to Crochet buddy’s handle. I then used this setup to look into how it would interpret three different stitches: the chain stitch, the double crochet, and the half double crochet.
Crochet stitches are generally fairly unique and have multiple distinct phases of how many loops of yarn are on the hook at a given time. Each stitch begins and ends with only one single loop of yarn on the hook, a very distinctive indicator when counting stitches. These features make Crochet Buddy the perfect them the perfect candidate for a system to artificially track the number of loops on the hook at any given time.
The first stitch that I looked at was the chain stitch which starts with a slip knot to create the first loop on the crochet hook. Next, you loop yarn around the hook so that there are two loops around the hook, and then pull the yarn through the first loop so that there is only one loop on the hook again. From the video footage taken, there seemed to be a very clear view of the one loop on the hook and the two loops on the hook which are the biggest indicators of this stitch.
The second stitch, the double crochet, which starts with a chain of however long you want the piece to be. At the end of this chain, there will be one loop on the hook, which is the base for this stitch. You then loop the yarn around the hook so that there are two loops on the hook. Then you push the hook through the stitch closest to the hook. There should now be three loops on the hook. You then loop the yarn around the hook so that there are four loops on the hook. Then, you pull the hook through two of the four stitches so that there are two loops on the hook. You then loop the yarn around the hook so that there are three loops on the hook. Then you pull the hook through two of the three loops so that there is only one loop on the hook, which completes the stitch. From the video footage taken, there seemed to be a very clear view of all the loops on the hooks.
The final stitch, the half double crochet, is fairly similar to the double crochet. Similar to the double crochet, which starts with a chain of however long you want the piece to be. At the end of this chain, there will be one loop on the hook, which is the base for this stitch. You then loop the yarn around the hook so that there are two loops on the hook. Then you push the hook through the stitch closest to the hook. There should now be three loops on the hook. You then loop the yarn around the hook so that there are four loops on the hook. Then, you pull the hook through three of the four loops on the hook so that there is only one loop on the hook again, completing the stitch. From the video footage taken, there seemed to be a very clear view of all the loops on the hooks used during this stitch making them usable in the AI process.
View the video of the Crochet Buddy Bodystorming here:
ANALYSIS
The analog crochet buddy prototype, coupled with my iPad positioned at a specific angle, was an incredible way to gain a deeper understanding of what the AI algorithm would be working with if it analyzed the video data from the camera positioned in the handle of crochet buddy. Although the video footage came out in a way that seemed to be appropriate for the AI tracking activity I had expected, there were two main edge cases that I came across which will be incredibly important to consider when designing to algorithms for Crochet Buddy’s stitch tracking.
While making the video, I also noticed that even though I am pretty decent at crochet, I found that I still occasionally mess up stitches. In the current set up of crochet buddy, there is no way to let crochet buddy know that you messed up a stitch, and it could cause great error in the tracking if you messed up a stitch and it still counted it, or tracked the type of stitch that it was incorrectly. In order to combat this, there needs to be some version of a command which let’s Crochet Buddy the user messed up a stitch. Either through an “I messed up the stitch” button, which will go back to the most recent single loop on the crochet hook, or through a voice command such as “Go back to previous stitch,” which would prompt the AI algorithm to go back to the spot in the video with the last correctly completed stitch.
I also noticed that during the video footage, my finger occasionally passed in front of the camera. Although it did not seem to have an effect on being able to view the see the number of loops on the hook. But, that does not mean that there isn’t a possibility for someone's fingers to genuinely block the view of the camera. In order to track stitches correctly, there needs to be extensive testing with multiple types of fingers to ensure that there will not be any miss tracked stitches due to a user’s large fingers.
